BUG: Order find results by score inside the search region.

[breken-ai]

Fixes #70

Description

find / find_until / get_element_coords crash with IndexError (or rank matches by the wrong score) whenever the search region does not start at the top-left corner of the screen, for example bot.find("label", x=100).

locate_all_opencv sorts its matches with key=lambda p: result[p[1]][p[0]], where p is the returned screen position. That position already includes the region offset (matches[1] * step + region[0]), but result (the cv2.matchTemplate output) is indexed from the region's own origin. So the key reads the score of a different pixel, and once x + region[0] passes the end of result it raises:

IndexError: index 110 is out of bounds for axis 0 with size 91

The sort was added in #25 ("Order find results by best match."); any x or y offset has hit this path since then.

Fix

Take the scores at the match positions inside result (result[matches]), sort by them, and keep adding the offset only for the returned Box. Behaviour with no offset is unchanged.

Tests

New botcity/core/tests/test_cv2find.py, pure numpy images, no real screen:

• locate_all_opencv(..., region=(100, 0, 100, 100)) returns the match at (150, 60)
• with two matches in an offset region, the exact copy comes before the damaged copy
• DesktopBot.find("needle", x=100) with get_screenshot / _fix_display_size stubbed returns the element

On main all three fail with the IndexError above; with this change python run_tests.py passes (4 passed) and flake8 botcity is clean.

---

This PR was prepared by an AI coding agent (Breken) running as the breken-ai account. I checked the change and the tests before opening it and am happy to change anything.